Werk #22290: Redfish: keep last good data while a resource is not ready
| Component | Checks & agents | ||||||
| Title | Redfish: keep last good data while a resource is not ready | ||||||
| Date | Sep 24, 2026 | ||||||
| Level | Trivial Change | ||||||
| Class | Bug Fix | ||||||
| Compatibility | Compatible - no manual interaction needed | ||||||
| Checkmk versions & editions |
|
Some management controllers, notably HPE iLO, intermittently respond to
requests for individual resources with a temporary-unavailable error,
e.g. a storage controller with HTTP 400
iLO.2.25.ResourceNotReadyRetry. The special agent omitted that entry
from its output but still reported success, so the storage controller,
its drives and its volumes could briefly show UNKNOWN with Item not
found in monitoring data.
The agent now recognises such responses (HTTP 503,
ServiceTemporarilyUnavailable, ResourceNotReadyRetry) and leaves the
affected section and the sections depending on it out of its output.
Requests that fail otherwise, e.g. with a timeout, are treated the same
way. Checkmk then keeps using the last successfully received data of
these sections for up to five minutes, so this helps if the Check_MK
service of the host runs at least every five minutes. Sections the
device returns empty are sent as empty, so removed hardware does not
linger.
As a side effect, services using persisted Redfish sections are now shown as based on cached agent data. If the service root, an advertised Manager collection or the Chassis collection is temporarily unavailable, the agent now aborts the run, so the previous data is kept and the Check_MK service turns CRIT. The system data retry introduced with Werk #19986 is unchanged.